Friday, September 6, 2019

Lessons Gleaned From The NTSB Report For A Tesla Autopilot-Engaged Car Crash

L. Elliot, Sept 4, "The NTSB has released its findings about a car crash that involved a Tesla Model S with its Autopilot-engaged that rammed into the back of a parked fire truck on a busy freeway in Southern California, occurring on a sunny morning of January 22, 2019…."  Read more   Hmmm…  While the "Factual Report of the Investigation" provides the background and some interesting tidbits, the important information is contained in the:

  • "Vehicle Data Recorder Specialist’s Factual Findings": especially the charts of "Vehicle drive mode information": Figures 1 for the hour leading up to the crash and Figure 2, for just the 15 minutes prior.  It is very interesting to have the precision and richness of data of the vehicle’s behavior prior to the crash.  Armed with this information, no wonder Elon wants to insure these cars.   What is most interesting about these data is the chart of Lead Vehicle Distance (m).  It shows that "lead vehicle distance" is not the instantaneous value obtained by the radar but some smoothed out value of { previous readings plus the latest radar value} (else, there would be some discrete jumps in the data when other cars either cut-in or cut-out of the Tesla’s lane ahead.).  Moreover, the appearance of a stationary object (approach speed = Tesla speed) in the lane ahead is disregarded (or very lightly weighted) in the determination of "lead vehicle distance". (it grew to its saturated value (that was much greater than the distance to the firetruck) once the lead SUV had changed lanes (whenever that was determined to have occurred).  At some point (possibly 490msec before the crash, see below), the system decided that the stationary object detected ahead was not a "false reading" but actually a stationary object that should no longer be disregarded.  Since it was being disregarded the Traffic Aware Cruise Control (TACC) operated using a large value for "lead vehicle distance" so it began to accelerate to its desired cruise speed, as would be expected if "lead vehicle distance" is a large value.  Yipe!!!!!  If Elon hasn’t already demanded, NTSB should require Tesla, and all other manufacturers, to: 1.  The software/logic governing TACC’s behavior during  transitions involving a cut-out or a cut-in needs to be substantially improved!, and  2.  The reliability in the identification of stationary objects in the lane ahead needs to be substantially improved so that they cease to be assumed to be false alarms.   
  • "Vehicle Automation Data Summary Report": especially:

1.  Figure 4, The speed of the Tesla in the last 221 seconds before the crash showing that the Tesla was traveling rather slowly in the 100 seconds before the crash (under 20 mph), but then accelerated (as discussed above) in the 3 seconds just prior to the crash, beginning as soon as the lead SUV changed lanes,

2.  Figure 5,  the distance between the Tesla and its lead vehicle, showing that the TACC worked really well until the lead vehicle "disappeared" (changed lanes), and

3.  Figure 6  which clearly depicts the movement of the Tesla relative to the lead vehicle and the Firetruck in the 15 seconds before the crash.   The Tesla’s radar and front facing camera mush have "seen’ the firetruck 4 seconds before the crash and every sensing loop (1/10th of a second) during the last 4 seconds yet…

"… Data show that at about 490 msec before the crash, the system detected a stationary object in path of the Tesla. At that time, the forward collision warning was activated; the system presented a visual and auditory warning. Data also shows that the AEB did not engage and that there was no driver-applied braking of steering prior to the crash. According to Tesla, the AEB was active at the time of the crash, and considering that the stopped fire truck was detected about half a second before impact, there likely was not sufficient time to activate the AEB."  …This implies that the AEB and its functioning in collaboration with the TACC needs to be substantially re-evaluated/re-designed.  Alain

   Smart Driving Cars Podcast Episode 123 – K. Kolodge JD Power

F. Fishkin, Aug 30 , "A J. D. Power study finds customer demand for safety technology threatened by overbearing alerts. Lead researcher Kristin Kolodge joins Princeton’s Alain Kornhauser and co-host Fred Fishkin for that plus headlines from Tesla, NVIDIA, GM’s Cruise, Lyft and Ford.  "  Just say "Alexa, play the Smart Driving Cars podcast!".  Ditto with Siri, and GooglePlay …  Alain


  Customer Demand for Safety Technology Threatened by Overbearing Alerts, J.D. Power Finds

Press Release, Aug 27, "Some alerts on Advanced Driver Assistance Systems (ADAS) are so annoying or bothersome that many drivers disable the systems and may try to avoid them on future vehicle purchases, according to the J.D. Power 2019 U.S. Tech Experience Index (TXI) Study,SM released today. This is a major concern for automakers keen to market these lucrative technologies and pave the way for more highly automated vehicles in the future.

"Automakers are spending lots of money on advanced technology development, but the constant alerts can confuse and frustrate drivers," said Kristin Kolodge, Executive Director of Driver Interaction & Human Machine Interface Research at J.D. Power. "The technology can’t come across as a nagging parent; no one wants to be constantly told they aren’t driving correctly."

A prime example of this is lane-keeping and centering systems. On average, 23% of customers with these systems complain that the alerts are annoying or bothersome. This ranges from just 8% for one domestic brand to more than 30% for a couple of import brands. For these owners, 61% sometimes disable the system, compared with just 21% of those that don’t consider the alerts annoying or bothersome. Owners wanting the feature on their next vehicle ranges from 63% for those that consider the alerts annoying or bothersome to 91% for those who do not. …"  Read more   Hmmm…  This highlights the fundamental problem with "warning systems" as opposed to "doing systems".   " Warning systems" are "warning systems" because they aren’t good enough at being "doing systems".  If they were, they’d be "doing" instead of just "Hey, Yo!, something screwy may be going on ahead.  I’m not sure if you’re paying attention so maybe you should start paying attention.  Just sayin’, I’m not sure of what to do, hope you know!" 

Lane keeping systems should be keeping you in your lane and be good enough to "cheat a little" if "cheating a little" is both available and necessary.  If it is not available and necessary, then the bakes need to be applied.  If it is not necessary and available, let the driver be the driver.  If the system isn’t good enough to determine necessity, then JD Power or some trusted "Good Housekeeping"  should rate it as poor/useless ’cause you’ll turn it off".

Unfortunately this study conflates driving technologies with comfort & convenience technologies.  Yes, driving technologies deliver comfort & convenience but their objective is primarily crash avoidance.  Those technologies should be clearly differentiated and rated separately from the other technologies, some of which, like Apple CarPlay and Android Auto may actually net out to detract from alert driving.    Alain

  Welcome to San Diego. Don’t Mind the Scooters.

E. Griffith, Sept 4, "The first thing you notice in San Diego’s historic Gaslamp Quarter is not the brick sidewalks, the rows of bars and the roving gaggles of bachelorette parties and conferencegoers, or even the actual gas lamps.

It’s the electric rental scooters. Hundreds are scattered around the sidewalks, clustered in newly painted corrals on the street and piled up in the gutters. In early July, one corner alone had 37. In the area around Mission Beach, one of the city’s main beaches, a single side of one block had 70. Most sat unused.

Since scooter rental companies like Bird, Lime, Razor, Lyft and Uber-owned Jump moved into San Diego last year, inflating the city’s scooter population to as many as 40,000 by some estimates, the vehicles have led to injuries, deaths, lawsuits and vandals. Regulators and local activists have pushed back against them. One company has even started collecting the vehicles to help keep the sidewalks clear. …

Safety has become a big issue. A three-month study published in May from the Centers for Disease Control and Prevention and the Public Health and Transportation Departments of Austin, Tex., found that for every 100,000 scooter rides, 20 people were injured. Nearly half of the injuries were to the head; 15 percent of those showed evidence of traumatic brain injury….

San Diego initially took a hands-off approach. The scooters became popular, with an average of 30,000 riders per day, according to city officials….

In July 2018, he teamed up with John Heinkel, owner of a local towing company, to haul away scooters that they deemed to be parked on private property. They charge Bird, Lime and others a retrieval fee of $50 per scooter, plus $2 for each day of storage….Their company, ScootScoop, has essentially turned them into scooter bounty hunters. They said they have struck deals with 250 local businesses and hotels and have towed more than 12,500 scooters….

In December, a man in Chula Vista, a San Diego suburb, died after he was hit by a car while riding a Bird scooter, according to the Chula Vista Police Department. A tourist died a few months later after crashing his rental scooter into a tree. Another visitor died of “blunt force torso trauma” after his scooter collided with another, the San Diego Police Department said.

The department said it counted 15 “serious injury collisions” involving scooters in the first half of this year. Last month, three separate scooter-related skull fractures happened in one week…."  Read more   Hmmm… Scooters have advocates, customers and investors, but if they aren’t "welcomed" in the neighborhoods they traverse they’ll at best be a hula-hoop craze and vanish.   Driverless Mobility-as-a-Service could/will face similar insurmountable challenges if it doesn’t earn, before starting services, a "welcoming" by each and every community that it traverses in delivering its Mobility-as-a-Service. Doing an "Uber"/"Scooter" (we’re doing this , like it or not, so lawyer-up!) isn’t going to work for Driverless Mobility-as-a-Service any better than it worked for "Uber" or "Scooter".  Good to start, pick the pockets of investors and then get completely wiped.  Alain

  Tesla Just Made a Huge Announcement That May Completely Change the Auto Industry. Here’s Why It’s Brilliant

J. Bariso, Sept. 3, "… "Starting today, we’re launching Tesla Insurance, a competitively priced insurance offering designed to provide Tesla owners with up to 20 percent lower rates, and in some cases as much as 30 percent," Tesla announced on its website last week. According to the statement, Tesla Insurance offers comprehensive coverage and claims management to customers in California, with planned expansion to additional U.S. states in the future.

The insurance problem is one Musk has been itching to solve for some time. Tesla vehicles are notoriously expensive to insure, despite the company’s claims that its cars are the safest in the world.

Analysts claim the reasons Teslas are so expensive to repair has to do with the cars’ aluminum construction, limited expertise on the part of mechanics and auto-body repair experts, and cost of replacement parts….

Instead, all indications are that Tesla’s plan is quite brilliant …. Through all of this, the company gets to leverage a major strength, one of Tesla owners’ favorite things about the company:

Its customer service.

Of course, if Tesla can prove that it will carry that same customer service over to the insurance side, it will gain consumer trust and loyalty. …."  Read more  Hmmmm… What is most important about the data collected by Tesla, is that it can readily "prove" who is at fault.  AutoPilot is focused on reducing crashes, so if a  fender bender/crash does occur, the likelihood that it is Tesla’s fault is reduced, plus the data allows Tesla to inexpensively determine who is at fault, thus allowing them to more cheaply pay, if at fault, or make the other guy pay if the Tesla is not at fault.  That’s what is really brilliant about this.  This may be so brilliant that it pays Elon to make AutoPilot even better, and certainly allows him to benefit from the AutoPilot beyond its initial selling revenue.  By improving AutoPilot and disseminating the improved software using over-the-air updates, he gets to keep forever more of the insurance revenue. That is real customer service… "Buy AutoPilot now and save on insurance for as long as you own the car."  Or, "Buy insurance from me and I’ll turn on AutoPilot for free."  (If you buy from the other the Gecko of Flo then they’ll have to pay me to turn on AutoPilot.) …  Plus fee updates, because we’ll get to keep even more of the insurance premium.  Life is good!!  🙂  See also Tesla Stakes Insurance Claim  by Roger Lanctot, Aug 23 Alain

  After Levandowski’s Indictment, Will Silicon Valley See the Error of Its Ways?

E. Niedermeyer, Sept 3, "Anthony Levandowski is back in the news after being indicted on 33 criminal counts related to his alleged theft of autonomous drive technology secrets from his former employer Google. Though the indictment doesn’t reveal much new information that didn’t come out in the 2017 Waymo vs. Uber civil lawsuit, it has once again thrust the most infamous man in autonomy into a much brighter spotlight than individuals in the close-knit AD community are accustomed to. But as easy as it is to dismiss Levandowski as a uniquely grotesque villain, his story calls as much for self-reflection in Silicon Valley as moralizing….

Buried in Silicon Valley’s lionization of "the crazy ones," established as high-tech orthodoxy by Apple’s memorable "think different" ad, is an implicit ends-justify-the-means morality that has simmered even as the tech sector has accumulated unprecedented financial and cultural clout. The logic isn’t even unique to The Valley, as badly-behaved rainmakers have enjoyed protection and prestige everywhere from Wall Street to Washington DC to Hollywood, but the technology business has gone farther in embracing and normalizing an explicit tradeoff between genius and rectitude than anywhere else….

With few remaining doubts that autonomous science projects can share public roads, AV development from here on out is fundamentally a pursuit of safety and trust. …" Read more  Hmmmm… "AV development from here on out is fundamentally a pursuit of safety and trust.FUNDAMENTAL!!! Amen!  Alain

  NVIDIA DRIVE LABS  Inside look at autonomous vehicle software

Staff, Aug. 23, "The DRIVE Labs video series takes an engineering-focused look at a range of self-driving challenges, from perceiving paths to handling intersections. These short clips illustrate how the NVIDIA DRIVE™ AV Software team is creating safe and robust self-driving systems…."  Read more   Hmmm…  See the various videos.  What is obvious from the "Ride in NVDIA’s Self-Driving Car" is that the system is disregarding stationary objects in the lane ahead. else it would be labeling overpasses and the clearance available under the overpass as opposed to assuming that sufficient clearance exists.  It would be identifying not only the lane ahead but also the height clearance envelope ("clearance tunnel")  ahead that will permit the  whole car to pass through and not just a sliver along the surface of the roadway.   That’s why this is "Do it yourSelf-driving" and absolutely requires an alert licensed driver behind the wheel, as is clearly proclaimed, (and NOT Driverless Mobility-as-a-Service). Alain

  Will Self-Driving Cars Really Only Last Four Years?

L. Elliot, Sept 3, "A recent news item about self-driving cars has been creating quite a flurry of debate due to the suggestion by a major automotive maker that driverless cars might only last about four years….


According to various stats about today’s cars, the average age of a conventional car in the United States is estimated at 11.6 years old….


Indeed, one of the fastest growing segments of car ages is the group that is 16 years or older, amounting to an estimated 81 million such cars by the year 2021. Of those 81 million cars, around one-fourth are going to be more than 25 years old. ….


When you buy a new car, the rule-of-thumb often quoted by automakers is that the car should last about 8 years or 150,000 miles.  This is obviously a low-ball kind of posturing, trying to set expectations so that car buyers will be pleased if their cars last longer. One supposes it also perhaps gets buyers into the mental mode of considering buying their next car in about 8 years or so….


Americans drive their cars for about 11,000 miles per year. If a new car is supposed to last for 150,000 miles, … you could drive the car for 14 years … so that is the situation with "Do-it-yourSelf-driving Cars.  Basically the same as today.  Except, since they deliver, some of the time, the Comfort&Convenience of Self-driving folks way move to cheaper land farther from where they work and go to events farther away.  The 11,000 miles per year average goes to 12, 15, …? k.  So "Do-it-yourSelf-driving car sales go up 10%, 30%, …?%.  Ya Hooooo for the OEMs!!!….


Various published stats about ridesharing drivers such as Uber and Lyft suggest that they are amassing about 1,000 miles per week on their cars. …  50,000 miles per year, ….  these on-the-go cars … Driverless Mobility-as-a-Service Cars…  would only last about 3 years,  … If land-use, mobility patterns don’t change, and there is no ride sharing, then you need fewer cars, but they are replaced more often, so that is awash at, say, 16M/year.  But more people have access to good mobility, so OEMs sell a few more  Driverless Mobility-as-a-Service Cars per year, maybe 18M/year.  If they are affordable, then vehicleTrip lengths (= personTrip length, since no ride-sharing) will increase.  OEMs will sell even more than the nominal.  Maybe 20M/year. Yea for the OEMs! 


If there is ride-sharing, rides become even more affordable.  personTrips & Length will increase, but vehicleTrips & total vehicleTripLength will decrease becasue more than one personTrip is served by each vehicle, on average.  Goodby 20M/year, even 16M/year.  Maybe as low as 10 or 12M/year. …"  Read more  Hmmmm… Whoops!!  OEMs don’t like ride-sharing Driverless Mobility-as-a-Service!   Alain

  Community Newsletter

Staff, Sept 3, "As August temperatures climbed in San Francisco, so did the miles driven by Cruise’s all-electric, self-driving cars.  In addition to testing in temperatures as high as 90 degrees, this month Cruise: … "  Read more  Hmmmm…  Interesting to see what Cruise is doing is SF with Community relations.  They certainly highlight the fact that they use EVs but I could not find a mention, nor any interest in casual ride-sharing, which is THE answer to "congestion, energy and pollution".  There is an excellent claim in one of the articles …"Cruise works closely with many neighborhood and merchant associations to better understand San Francisco’s transportation needs block by block so that we’ll be able to help small businesses reach more customers and connect more people to the places they want to go….".  Alain


Metro Labs, Aug 2019, "… As an extension of the Foundation’s work in the Greater Seattle region, the Community and Civic Engagements (CCE) Team set out to learn more about the use and impact of data and technology on civic and community priorities. In particular, the team was interested in exploring models that inspire and enable civic innovation and participation to develop a healthy and vibrant community.  In the spring of 2018, MetroLab Network kicked-off a year-long effort with the CCE team to help inform their learning process. That process included six sight visits to leading cities in civic innovation, data, technology, and engagement. The visits focused on a range of themes, including:
• Economic Mobility and Human Services
• Built and Natural Environment
• Civic Engagement
• Transportation and Public Services
• Data
• Institutional Partnerships

….  There has been a groundswell of interest in driving change and progress in communities by leveraging data, technology, and civic engagement. Such efforts take advantage of numerous technological advancements, including cloud storage and computing, sensor technology, machine learning and artificial intelligence, next-generation connectivity, all while considering social and behavioral dynamics in communities. As each community defines their vision of civic innovation and engagement, their work will serve as “laboratories of innovation,” serving as guides for the approaches that, if successful, will be adopted at greater scale. Through this exercise, The Gates Foundation and MetroLab were struck by the commitment and passion from communities seeking to leverage civic technology and innovation to drive social equity, policy priorities, and service delivery.
There are a number of take-aways that emerged from the site visits:
1. The secret to effective civic innovation ecosystems is more about a network of human
relationships than data- or technology approaches. …" Read more  Hmmmm… Very interesting 1st conclusion that human relationship/interaction is dominant.  This is really important for the proper launch and evolution of the SmartDrivingCar r/evolution. Alain

  The Human Cost of Amazon’s Fast, Free Shipping

P. Callahan, Sept 5, "… In its relentless push for e-commerce dominance, Amazon has built a huge logistics operation in recent years to get more goods to customers’ homes in less and less time. As it moves to reduce its reliance on legacy carriers like United Parcel Service, the retailer has created a network of contractors across the country that allows the company to expand and shrink the delivery force as needed, while avoiding the costs of taking on permanent employees.

But Amazon’s promise of speedy delivery has come at a price, one largely hidden from public view. An investigation by ProPublica identified more than 60 accidents since June 2015 involving Amazon delivery contractors that resulted in serious injuries, including 10 deaths. That tally is most likely a fraction of the accidents that have occurred: Many people don’t sue, and those who do can’t always tell when Amazon is involved, court records, police reports and news accounts show….  "   Read more  Hmmmm… How quickly does Amazon want to have these deliveries made by Driverless trucks in the wee hours of the morning when no one else is using the roads….  No one to hurt if there should be a crash and no drivers to manage.  Just Moore’s Law-ish algorithms, data, sensors and actuators.  Shipping will then cost almost nothing; therefore readily supported by the Prime membership fee.  Alain

  14 women have filed a lawsuit against Lyft accusing the company of not addressing a ‘sexual predator crisis’ among drivers

G. Rapier, Sept 5, "Fourteen victims of sexual assault have filed a lawsuit against Lyft accusing the company of failing to respond adequately to what they call a "sexual predator crisis" among drivers on its platform.

"Complaints to Lyft by female customers who have been attacked by Lyft drivers, combined with subsequent criminal investigations by law enforcement, …

The lawsuit also notes that many Lyft drivers have installed cameras in their cars to prevent — or help after the fact — any attacks from riders. One particular attack on a driver in New York earlier this year went viral after the driver shared it with local news outlets. Many other drivers speaking to Business Insider have outlined similarly harrowing incidents. …"  Read more  Hmmmm… We must behave better as a society, else we cease to be a community.  We must be sure that these types of issues are absent from Driverless Mobility-as-a-Service; else, ride-sharing have no-chance, and the whole Driverless Mobility-as-a-Service concept will be DoA.  Appropriately addressing this issue is a game-changing socio-technological challenge that, to date, has received essentially zero intellectual attention.  Alain

Tesla Model 3 becomes one of the best-selling cars in the UK — not just electric

F. Lambert, Sept 5,  "The ramp-up has been slow over the first few months, but now the registration numbers are out for August, and Tesla delivered more than 2,000 Model 3 vehicles that month.

It was enough for the Model 3 to become the third best-selling car in the UK in August (via the Society of Motor Manufacturers and Traders (SMMT) thanks to Mark Vandersluis): …"  Read more   Hmmmm…   Impressive!.  Alain

Half-baked stuff that probably doesn’t deserve your time

  Self-driving cars could ease future evacuations

Staff, Sept 4, "AVs might help make evacuations more efficient, former Florida emergency management chief Bryan Koon, now a vice president at disaster consulting firm IEM, wrote in a blog…"  Read more  Hmmmm…Totally Half-baked.  The Self-driving features are USELSS in evacuation. People are available.  They are being evacuated!  In this Operational Design Domain, Driverless is simply C’Mon Man!   Alain  

 C’mon Man!  (These folks didn’t get/read the memo)

Simply ClickBait

 Calendar of Upcoming Events:

4th Annual Princeton SmartDrivingCar Summit

evening May 19 through May 21, 2020

On the More Technical Side